lasso.boost {timereg} | R Documentation |
Fits the LASSO estimator for the additive risk model based on the least squares fitting criterion
L(β,D,d) = β^T D β - 2 β^T d
where D=int Z H Z dt and d=int Z H dN.
lasso.boost(D,d,lambda,max.it=10000,beta=0,detail=0)
D |
defined above |
d |
defined above |
lambda |
l1 regularization |
max.it |
number of steps in l1 boosting algorithm |
detail |
prints details |
beta |
starting value for algorithm. |
lasso.boost is the boosting algorithm
lasso.add.hazard computes the exact solution using the quadprog package.
returns a list with the following arguments:
beta |
regression coefficients |
L |
value of the fitting criterion |
l1 |
sum of the absolute value of the coeffcients |
Thomas Scheike
Martinussen and Scheike, The Aalen additive hazards model with high-dimensional regressors, submitted.
Martinussen and Scheike, Dynamic Regression Models for Survival Data, Springer (2006).
Kim and Kim (2004), Gradient LASSO for feature selection, In Proceedings of the 21st International Conference on Machine Learning.
## makes data for pbc complete case data(mypbc) pbc<-mypbc pbc$time<-pbc$time+runif(418)*0.1; pbc$time<-pbc$time/365 pbc<-subset(pbc,complete.cases(pbc)); covs<-as.matrix(pbc[,-c(1:3,6)]) covs<-cbind(covs[,c(1:6,16)],log(covs[,7:15])) ## computes the matrices needed for the least squares criterion out2<-aalen.test(Surv(time,status>=1)~const(covs),pbc,robust=0,n.sim=0) S=out2$intZHZ; s=out2$intZHdN; ## lambda=0.39 out.pbc<-lasso.boost(S,s,0.39,max.it=20000) ## exact solution, slow sometimes !! # library(quadprog) # out.ex<-lasso.add.hazard(S,s,0.39,0,1,max.it=20000) # print(round(cbind(out.pbc$beta,out.ex$sol$solution),4))